Pandigital arbitrary-base slow

#include <stdio.h>

int main(void)
{
    unsigned char q[21], r[21], t[21];
    for (int base=4; base<20; base+=2)
    {
        for (int i=0; i<20; i++) q[i]=0;
        q[base]=77;
        unsigned long long ct = 0;
        while (q[base]!=78)
        {
            for (int i=0; i<base; i++) r[i]=i;
            // convert from mixed-base to factorial
            for (int i=0; i<base; i++)
            {
                t[i]=r[q[i]];
                for (int j=q[i]+1; j<base; j++) r[j-1]=r[j];
            }
            // prepare the numbers
            unsigned long long Z1=0,Z2=0,Z3=0;
            for (int i=0; i<base/2; i++)
            {
                Z1=(Z1*base)+t[i];
                Z2=(Z2*base)+t[i+base/2];
            }
            if (Z1<Z2)
            {
                // multiply
                Z3 = Z1*Z2; unsigned long long Z3u=Z3;
                unsigned long long msk = 0;
                for (int i=0; i<base; i++)
                {
                    msk |= 1<<(Z3%base);
                    Z3/=base;
                }
                if (msk == (1<<base)-1)
                {
                    ct++;
                    // printf("%d  %llu %llu %llu\n", base, Z1, Z2, Z3u);
                }
            }

            // (base 4) 30*21=1230

            // and increment
            q[0]++;
            int j = 0;
            while (q[j]==(base-j) && j<base) { q[j]=0; q[1+j]++; j++; }
        }
        printf("%d  %d\n", base, ct);
    }
}